Great to see this proceeding! Thought I'd weigh in on this part: > Unfortunately, we had to introduce a new bind flag (XDP_USE_SG) on the > AF_XDP level to enable multi-buffer support. It would be great if you > have ideas on how to get rid of it. The reason we need to > differentiate between non multi-buffer and multi-buffer is the > behaviour when the kernel gets a packet that is larger than the frame > size. Without multi-buffer, this packet is dropped and marked in the > stats. With multi-buffer on, we want to split it up into multiple > frames instead. > > At the start, we thought that riding on the .frags section name of > the XDP program was a good idea. You do not have to introduce yet > another flag and all AF_XDP users must load an XDP program anyway > to get any traffic up to the socket, so why not just say that the XDP > program decides if the AF_XDP socket should get multi-buffer packets > or not? The problem is that we can create an AF_XDP socket that is Tx > only and that works without having to load an XDP program at > all. Another problem is that the XDP program might change during the > execution, so we would have to check this for every single packet. I agree that it's better to tie the enabling of this to a socket flag instead of to the XDP program, for a couple of reasons: - The XDP program can, as you say, be changed, but it can also be shared between several sockets in a single XSK, so this really needs to be tied to the socket. - The XDP program is often installed implicitly by libxdp, in which case the program can't really set the flag on the program itself. There's a related question of whether the frags flag on the XDP program should be a prerequisite for enabling it at the socket? I think probably it should, right? Also, related to the first point above, how does the driver respond to two different sockets being attached to the same device with two different values of the flag? (As you can probably tell I didn't look at the details of the implementation...) -Toke
